Eighteen AI stories landed between May 18 and 24. A jury verdict closed a two-year legal saga, Anthropic made two moves in one week, and Meta laid off 8,000 people while logging their keystrokes to train AI. Relve, a trusted AI intelligence platform, filtered through all of it. Here is what made noise, what to watch, and why most of it does not change what you ship this week.
Key Takeaways
- Musk lost the OpenAI lawsuit. Nine jurors took under two hours to reject all claims on statute of limitations grounds; OpenAI’s IPO path is now clear
- Anthropic acquired Stainless and shut down access for OpenAI, Google, and Cloudflare. The SDK infrastructure most AI labs depended on is now exclusively Anthropic’s
- Meta cut 8,000 jobs while simultaneously logging employee keystrokes to train AI models via its Model Capability Initiative
- Nvidia launched the Vera CPU, claiming a brand new $200 billion TAM with $20 billion in sales already recorded in 2026
- Google appealed its search monopoly ruling, targeting the data-sharing remedy that would give OpenAI and Perplexity court-ordered access to its search index
Nine jurors unanimously rejected Musk’s lawsuit against OpenAI, Altman, Brockman, and Microsoft on statute of limitations grounds. The jury found Musk should have known about the alleged harms before the legal deadlines, ranging from August to November 2021.
The underlying question of whether OpenAI violated its charitable mission was never ruled on. Judge Gonzalez Rogers said she had been prepared to dismiss the case herself. Musk announced an appeal to the Ninth Circuit. Noise until there is a ruling on appeal.
The trial also surfaced that Musk directed OpenAI researchers including Ilya Sutskever and Andrej Karpathy to work at Tesla without reimbursement in 2017, a detail that complicated his central argument.
Meta laid off 8,000 employees and reassigned 7,000 more to AI model training work. Termination emails arrived at 4am. Engineer David Frenk produced a parody video targeting Meta’s Model Capability Initiative, a tool that logs employee mouse clicks and keystrokes to train AI models.
The video spread internally and on Blind. Workers subject to this monitoring were simultaneously being evaluated for redundancy. Meta has not publicly addressed the backlash.
The layoff pattern is consistent with Cloudflare, Oracle, and PayPal from prior weeks. Headcount cuts framed as AI transformation while infrastructure costs drive the actual decision.
Spotify launched ElevenLabs-powered audiobook creation in invite-only beta for June, Studio by Spotify Labs for personalised on-demand podcast generation from email and calendar data, and in-episode AI Q&A for Premium users in the US, Sweden, and Ireland.
Amazon’s Alexa+ now generates on-demand podcast episodes from any topic a user names, with content partnerships across 200-plus news publishers.
Stability AI released Stable Audio 3.0, generating compositions up to six minutes twenty seconds, trained on fully licensed data via Universal Music Group and Warner Music Group partnerships. Three audio platforms, one week. None ready for production use at scale today.
Figma launched a native AI agent inside its collaborative canvas, fine-tuned for design context, alongside its existing Code to Canvas integration with Claude Code. Q1 2026 revenue was $333.4 million, up 46% year on year. Stock is down roughly 85% from its 52-week high. Catch-up move in a market where Canva, Adobe, and Dessn are already moving.
DeepSeek made its 75% V4-Pro price cut permanent, dropping API costs to 0.025 to 6 yuan per million tokens. ByteDance, Tencent, and Alibaba are all scrambling to secure Huawei Ascend 950 chip orders. Supply remains constrained with Huawei targeting 750,000 units this year.
SandboxAQ integrated its drug discovery and materials science models into Claude via natural language interface. The integration removes the need for users to supply their own computing infrastructure. The interface change is real. The scientific workload is not reduced. Experimental validation is still required.
IBM rebuilt the Ferrari fan app with watsonx AI, reporting a 62% engagement lift over race weekends with no baseline or independent verification. The fan app metric is marketing copy.
The more useful data point: Lightspeed Venture Partners formalised a program using F1 paddock access to introduce portfolio founders to CIOs and CISOs at enterprise buying events. Oracle, Microsoft, CoreWeave, and Anthropic all sponsor F1 teams. The paddock is now a deal venue.
Why we’re watching these. These stories did not produce a product change or an immediate action item. But the patterns they are setting will shape AI infrastructure ownership, governance, and competitive dynamics before the year is out.
Anthropic acquired Stainless, the SDK automation startup that built developer libraries for OpenAI, Google DeepMind, Cloudflare, Replicate, Runway, Nvidia, and Perplexity. All hosted Stainless products are wound down effective immediately with no new signups, projects, or SDKs available.
Existing customers retain rights to SDKs already generated. Roughly a quarter of the world’s professional software developers have used an SDK or documentation site built with Stainless tools.
Competitors must now rebuild SDK automation in-house, procure alternatives, or maintain existing codebases manually as their APIs evolve. Companies relying on Stainless-managed SDKs should visit app.stainless.com/transition immediately.
Andrej Karpathy joined Anthropic’s pre-training team, reporting to team lead Nick Joseph, to build a group focused on using Claude to accelerate its own pre-training research. Karpathy co-founded OpenAI, led Tesla’s Full Self-Driving program, and most recently founded Eureka Labs.
Separately, Chris Rohlf, a 20-year cybersecurity veteran from Meta and Yahoo’s The Paranoids, joined Anthropic’s Frontier Red Team. The two hires signal where Anthropic is concentrating resources ahead of its next model generation: deeper pre-training capability and more rigorous adversarial safety testing.
Nvidia launched Vera, a CPU purpose-built for agentic AI token throughput, with $20 billion in standalone sales already recorded in 2026 and a claimed $200 billion TAM. Nvidia posted $81.6 billion in Q1 revenue and guided $91 billion for Q2.
Early adopters include Meta, Alibaba Cloud, ByteDance, Oracle Cloud Infrastructure, CoreWeave, and Cloudflare. AWS, Google, and Microsoft are all running parallel chip programs. The $200 billion claim assumes Nvidia wins the agent CPU market before those programs mature. Watch the enterprise standardisation decision, not the launch announcement.
SpaceX’s S-1 filing revealed xAI lost $6.4 billion from operations in 2025 on $3.2 billion in revenue. AI capex hit $12.7 billion for the year and accelerated to $7.7 billion in Q1 2026 alone.
Anthropic is paying $1.25 billion per month through May 2029 for compute access at Colossus and Colossus II. The NAACP filed a lawsuit over 46 unregulated mobile gas turbines at xAI’s Memphis data centers.
SpaceX’s S-1 plans to buy $2.8 billion more in turbines over three years despite the active litigation. Grok has 117 million monthly active AI users. SpaceX targets a $1.75 trillion valuation on Nasdaq under “SPCX.”
Google filed its appeal brief on May 22, challenging Judge Mehta’s 2024 ruling that found it held illegal search monopolies.
The most consequential section targets the data-sharing remedy that would give OpenAI and Perplexity court-ordered access to Google’s search index. If overturned, AI search entrants compete without that structural advantage. DOJ response expected July 2026.
Trump delayed signing the AI security executive order on May 21, citing language he said could have been a blocker for US AI leadership over China. The EO would have required AI companies to share frontier models with the government for 14 to 90 days before public release. No formal pre-release review framework currently exists.
Google Search went AI-first at I/O 2026 the same week. Kagi, DuckDuckGo, Startpage, and Brave are the main alternatives for teams that want out.
Missed last week’s roundup? Catch up here.
The announcements that actually change how AI is deployed inside enterprise teams are a different category. Browse all recent signals at Relve. The Company OS series published this week covers what the meeting intelligence shift means for Engineering, Ops, HR, and Marketing specifically.
Don’t sort through the noise yourself every week. Subscribe to Relve and get the signal briefing that actually matters, free every week.
